The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/office/soffice.py contains logic to dynamically generate C source code, compile it into a shared object (.so) using gcc, and then inject it into the soffice process using the LD_PRELOAD environment variable. This is intended to shim AF_UNIX socket calls in restricted environments.
[COMMAND_EXECUTION]: Multiple scripts execute arbitrary system commands via subprocess.run():
scripts/office/soffice.py calls gcc to compile the socket shim.
scripts/recalc.py calls soffice (LibreOffice) and gtimeout.
scripts/office/validators/redlining.py calls git diff to compare document versions.
[INDIRECT_PROMPT_INJECTION]: The skill represents a significant indirect prompt injection surface as it is designed to ingest and process untrusted spreadsheet and Office XML data from the user's environment.
Ingestion points: Spreadsheet files read via pandas.read_excel() and openpyxl.load_workbook() in SKILL.md and validator scripts.
Boundary markers: None detected in the provided code snippets to distinguish between data and instructions within the spreadsheet content.
Capability inventory: Includes file system writes, execution of subprocesses (soffice, git, gcc), and macro management.
Sanitization: Relies on third-party libraries like pandas and openpyxl for parsing, but does not implement explicit content filtering for instructions embedded in cell data.
[PERSISTENCE_MECHANISMS]: The script scripts/recalc.py automatically modifies the user's LibreOffice configuration by writing a macro file (Module1.xba) to ~/.config/libreoffice/ (Linux) or ~/Library/Application Support/LibreOffice/ (macOS). while functional, this modifies application behavior across sessions.

xlsx